Can You Learn a Language Without Vocabulary Lists?

The short answer is yes, and for most people reading this it is also the right answer. But the case for frequency lists is stronger than input enthusiasts usually admit, and it is strongest in exactly the place where reading is weakest — at the very beginning, before there is enough language in place for reading to be possible at all.

This article is a verdict rather than a mechanism. How words are picked up from text is a separate question with a separate answer; what follows assumes it and argues the other side.

What a frequency list is, and is not

A frequency list is a corpus ranked. Somebody counted the words in a large body of text — newspapers, novels, subtitles, transcripts, or a mixture — and put them in order. The commonest lists are the top 1,000, 2,000 or 5,000 of a language, sometimes as word families and sometimes as separate forms. The tradition is old: the vocabulary control behind the first graded readers came out of exactly this kind of counting in the 1920s and 1930s, and French had an official one, the français fondamental, by the 1950s.

What a list is not is a syllabus. Rank order tells you how often a word occurs in somebody else’s corpus. It says nothing about whether you can use the word, whether you will meet it this month, or whether it means the same thing in the sentences where it actually appears.

The arithmetic that makes lists attractive

The reason lists keep coming back is that the first stretch of one is an extraordinary bargain, and the rest of it is not.

Vocabulary is distributed so unevenly that roughly the commonest thousand word families cover something like three quarters of the running words of a general text. The second thousand adds perhaps five or six points. The third adds three or four. By the time you are learning the seventh thousand, each word buys a tiny fraction of a percentage point of coverage, and you are memorising words you will meet once a year.

So the curve itself contains the verdict. Deliberate list work is efficient at the bottom and inefficient everywhere else, and the crossover is somewhere around the second thousand — the point at which reading becomes possible and starts doing the job better.

The honest case for lists: the bootstrap problem

Reading builds vocabulary from context. Context is made of words. At two hundred words of the language you have no context, so nothing can be inferred from anything, and the loop that the whole method depends on cannot start.

This is the real argument, and it is not answerable by reading harder. A complete beginner cannot read comprehensibly, not because they lack tolerance for ambiguity but because there is no text in the language at their level; even the gentlest graded material assumes several hundred words. Getting through that first stretch by deliberate means — a list, a course, a phrasebook, an app, whatever is tolerable — is faster by a wide margin than waiting for it to accumulate. Three months of concentrated work on the first two thousand words is a realistic timetable. Acquiring the same two thousand from reading, if you could read, would take a year or more.

The list, on this argument, exists in order to end. Its purpose is to make itself unnecessary as quickly as possible.

What lists give that reading does not

• Speed at the bottom. As above, and it is the main one.
• Guaranteed coverage of a defined set. Reading leaves an idiosyncratic vocabulary shaped by whatever you happened to read. A learner who spent a year on detective stories knows the words for alibi and coroner and not the words for invoice and appointment. A list has no holes because somebody else chose the contents.
• A deadline. If a specific set of words is required by an examination in April, waiting to meet them incidentally is not a plan.
• A domain. A bounded list for a narrow purpose — the two hundred words of your own profession, the vocabulary of the particular book you are about to start — is a good tool at any level, because within the domain the frequency is high even when it is low in the language.

What lists do not give

Everything else, and it is worth being specific, because the gap between “I have studied 2,000 words” and “I have 2,000 words” is where most of the disappointment lives.

• Most of the word. A list entry is a form and a translation. That is one thin slice of knowing a word: not its collocations, not its register, not what it implies, not the grammar it takes with it.
• The other meanings. Frequent words are polysemous, and the more frequent they are the worse it gets. The list gives you one gloss for a word that does eight jobs, and the eight jobs are exactly why it is at the top of the list.
• The forms you will actually see. A list gives citation forms. In Russian that is a nominative singular and an infinitive, and the page will show you neither — a noun in one of six cases, a verb from an aspect pair whose partner is a different word. In French or Spanish the gap is smaller but the same in kind. Studying citation forms and expecting to read is a well-trodden way to know two thousand words and understand nothing.
• Retention. A word learned from a list and never met decays, and the decay is fast. Deliberate learning does not durably install a word; it front-loads it, and the encounters in real text are what pay it off. A list without reading behind it is a bucket with a hole in it.
• The right corpus. Rank order depends entirely on what was counted. Lists built from film subtitles are excellent for conversation and mediocre for literature; a spoken-corpus list will teach you the vocabulary of a phone call before it teaches you the vocabulary of a paragraph. If your goal is reading, check what the list was made of.

The cost that is usually left out

Time spent on lists is time not spent reading, and the exchange rate changes as you go up. At the beginning it is strongly favourable: an hour on the commonest words does more than an hour staring at a text you cannot decode. Somewhere past the second thousand it turns, and at that point continuing to grind lists is not merely inefficient — it is displacing the activity that would have taught you the same words better, along with everything about them that the list omits.

There is a recognisable failure mode here. The learner who has done six thousand words in a vocabulary app and cannot read a page is not unlucky; they have optimised a measurable proxy at the expense of the thing it was standing in for.

The verdict, by situation

• Absolute beginner, no reading possible yet. Use a list. Take the commonest 1,000 to 2,000, learn them by whatever method you will actually sustain, and start reading graded material as early as you can bear — in parallel, not afterwards. Expect to stop the list work within a few months.
• You can get through a graded reader. Stop. Your reading is now a better list than any published one: it delivers the words you actually meet, at the frequency they actually have for you, with the context, the collocations and the grammar attached. Adding a general frequency list at this point buys ranked words you would have met anyway.
• An examination or a professional domain. Use a bounded list for the bounded purpose, and do not let it grow into a general one.
• Starting a specific difficult book. Harvest the twenty or thirty words that book keeps repeating and learn those. This is the most useful list anyone can make and it takes half an hour — one of the standard remedies for a text above your level.
• A heavily inflected language. If you use a list at all, use one that shows the forms: gender with the noun, the aspect partner with the verb, the case a preposition governs. A bare citation form is worth much less here than the list implies.

The thing that replaces the list

Nothing about giving up frequency lists requires giving up deliberate vocabulary work. It changes where the words come from. Words harvested from your own reading are already ranked by the only frequency that matters to you, already carry a sentence you understood, and are already at the third or fourth encounter rather than the first, which is the point at which deliberate work is cheapest and sticks best.

That is a small, specific practice with a few rules of its own, and it is the subject of the next article.

How far the first two thousand words actually get you, in coverage of a real page, is a question with numbers attached — and the reason the numbers matter is the input hypothesis itself.